Methods › General › Stochastic Optimization › AdaDelta
AdaDelta
Introduced by Matthew D. Zeiler in ADADELTA: An Adaptive Learning Rate Method
archive 2025-07-28 Description, source and code snippet are the archive's method entry.
AdaDelta is a stochastic optimization technique that allows for per-dimension learning rate method for SGD. It is an extension of Adagrad that seeks to reduce its aggressive, monotonically decreasing learning rate. Instead of accumulating all past squared gradients, Adadelta restricts the window of accumulated past gradients to a fixed size w.
Instead of inefficiently storing w previous squared gradients, the sum of gradients is recursively defined as a decaying average of all past squared gradients. The running average E[g²]ₜ at time step t then depends only on the previous average and current gradient:
E[g²]ₜ = γE[g²]ₜ₋₁ + (1-γ)g²ₜ
Usually γ is set to around $0.9$. Rewriting SGD updates in terms of the parameter update vector:
Δθₜ = -η·g_(t, i) θₜ₊₁ = θₜ + Δθₜ
AdaDelta takes the form:
Δθₜ = -η/(√(E[g²]ₜ + ϵ))gₜ
The main advantage of AdaDelta is that we do not need to set a default learning rate.
Papers archive 2025-07-28
17 shown of 17, newest first. Repository counts are the archive's code-links table. A Syntology line states what Syntology ran from that paper's harvested code; it is per sample and not a correctness claim.
-
Accelerating Energy-Efficient Federated Learning in Cell-Free Networks with Adaptive Quantization 30 Dec 2024 · 0 repositories · arXiv:2412.20785
-
New Insight in Cervical Cancer Diagnosis Using Convolution Neural Network Architecture 23 Oct 2024 · 0 repositories · arXiv:2410.17735
-
A Parallelized, Adam-Based Solver for Reserve and Security Constrained AC Unit Commitment 10 Oct 2023 · 1 repository · arXiv:2310.06650
-
ELRA: Exponential learning rate adaption gradient descent optimization method 12 Sep 2023 · 0 repositories · arXiv:2309.06274
-
Occupant's Behavior and Emotion Based Indoor Environment's Illumination Regulation 19 Feb 2023 · 0 repositories · arXiv:2302.09586
-
BFE and AdaBFE: A New Approach in Learning Rate Automation for Stochastic Optimization 6 Jul 2022 · 0 repositories · arXiv:2207.02763
-
Gradient Descent, Stochastic Optimization, and Other Tales 2 May 2022 · 0 repositories · arXiv:2205.00832
-
AdaSmooth: An Adaptive Learning Rate Method based on Effective Ratio 2 Apr 2022 · 0 repositories · arXiv:2204.00825
-
Adaptively Customizing Activation Functions for Various Layers 17 Dec 2021 · 1 repository · arXiv:2112.09442
-
Tom: Leveraging trend of the observed gradients for faster convergence 7 Sep 2021 · 2 repositories · arXiv:2109.03820
-
Exploiting Adam-like Optimization Algorithms to Improve the Performance of Convolutional Neural Networks 26 Mar 2021 · 0 repositories · arXiv:2103.14689
-
An Adaptive and Momental Bound Method for Stochastic Learning 27 Oct 2019 · 2 repositories · arXiv:1910.12249Syntology ran 0 of 2 samples · 2 unverified
-
diffGrad: An Optimization Method for Convolutional Neural Networks 12 Sep 2019 · 1 repository · arXiv:1909.11015
-
On the Convergence of Adam and Beyond 19 Apr 2019 · 3 repositories · arXiv:1904.09237Syntology ran 0 of 2 samples · 2 unverified
-
Adaptive Methods for Nonconvex Optimization 1 Dec 2018 · 1 repository
-
Online Batch Selection for Faster Training of Neural Networks 19 Nov 2015 · 1 repository · arXiv:1511.06343
-
ADADELTA: An Adaptive Learning Rate Method 22 Dec 2012 · 12 repositories · arXiv:1212.5701Syntology ran 0 of 6 samples · 6 unverified
Tasks archive 2025-07-28
11 tasks the archive attaches to papers tagged with this method, by distinct papers. A task without a page in the catalog is plain text.
Usage over time archive 2025-07-28
Components: the archive holds no method-to-method composition, so PwC's Components table cannot be rebuilt; the Papers list carries no Results column for the same reason (the archive does not join its leaderboard rows to method tags).
Categories archive 2025-07-28
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections